Natural Hazards and Earth System Sciences (2002) 2: 57–72
c European Geophysical Society 2002
Natural Hazards
and Earth
System Sciences
A geomorphological approach to the estimation of landslide hazards
and risks in Umbria, Central Italy
M. Cardinali1 , P. Reichenbach1 , F. Guzzetti1 , F. Ardizzone1 , G. Antonini2 , M. Galli2 , M. Cacciano3 , M. Castellani3 ,
and P. Salvati3
1 CNR
– IRPI, via della Madonna Alta 126, 06128 Perugia, Italy
s.n.c. Servizi al territorio, viale Vittorio Veneto 14/A, 06050 Papiano, Italy
3 Geologists, private consultants, Perugia, Italy
2 Terra
Received: 8 October 2001 – Accepted: 9 November 2001
Abstract. We present a geomorphological method to evaluate landslide hazard and risk. The method is based on the
recognition of existing and past landslides, on the scrutiny
of the local geological and morphological setting, and on the
study of site-specific and historical information on past landslide events. For each study area a multi-temporal landslide
inventory map has been prepared through the interpretation
of various sets of stereoscopic aerial photographs taken over
the period 1941–1999, field mapping carried out in the years
2000 and 2001, and the critical review of site-specific investigations completed to solve local instability problems. The
multi-temporal landslide map portrays the distribution of the
existing and past landslides and their observed changes over
a period of about 60 years. Changes in the distribution and
pattern of landslides allow one to infer the possible evolution
of slopes, the most probable type of failures, and their expected frequency of occurrence and intensity. This information is used to evaluate landslide hazard, and to estimate the
associated risk. The methodology is not straightforward and
requires experienced geomorphologists, trained in the recognition and analysis of slope processes. Levels of landslide
hazard and risk are expressed using an index that conveys,
in a simple and compact format, information on the landslide frequency, the landslide intensity, and the likely damage
caused by the expected failure. The methodology was tested
in 79 towns, villages, and individual dwellings in the Umbria
Region of central Italy.
Key words. Landslide, Hazard, Risk, Geomorphology, Inventory map, Umbria
1 Introduction and background
Much of Italy consists of hilly and mountainous terrain subject to landslides. Investigations aimed at assessing landslide
risk in Italy have shown that in the last century landslides
Correspondence to: M. Cardinali
([email protected])
affected at least 23 600 sites. This is equivalent to a density (on average) of about one landslide per 10 km2 . In the
20th century at least 7799 landslide casualties, and at least
100 000 homeless or evacuated people were reported. Between 1990 and 1999 at least 263 people were killed by
mass-movements, an average of 26 deaths per year (Guzzetti,
2000).
Damage and fatalities have been caused by single catastrophic failures and by widespread landsliding. The largest
landslide catastrophe in Italy occurred at Vajont (Veneto Region), on 9 October 1963. At 18:30 h, 240 million cubic meters of rock detached from the western slopes of M. Toc and
fell into the Vajont lake. The single rockslide pushed the water against the villages of Casso and Erto, and over the Vajont
dam. A water wave overtopped the dam and destroyed the
town of Longarone, killing at least 1917 people (Catenacci,
1992). Another highly destructive landslide occurred on 13
December 1982, at Ancona (Marche Region). The single
slope failure involved the movement of 342 hectares of urban
and suburban land, damage to two hospitals and the Faculty
of Medicine at Ancona University, damage to or complete
destruction of 280 buildings with a total of 865 apartments,
displacement of the main railway and coastal road for more
than 2.5 km, one (indirect) death, and the evacuation of 3661
people (Crescenti, 1986; Catenacci, 1992). The economic
loss was estimated at US$ 700 million (Alexander, 1989).
Economic damage and fatalities are also caused by
widespread shallow and deep-seated failures, triggered by intense or prolonged rainfall, by snowmelt, or by earthquakes.
In the last decade events triggering thousands of landslides
occurred in the Po Basin on 3–5 November 1994 (Regione
Piemonte, 1998) and on 13–16 October 2000, causing 22 and
25 fatalities, respectively, and economic damage in excess
of several million Euro. Other catastrophic events occurred
on 19 June 1996 in Versilia (Tuscany Region, 14 casualties),
and on 5–6 May 1998, at Sarno and Quindici (Campania Region), when secondary lahars detached from the slopes of
Pizzo d’Alvano, killing 153 people.
The Sarno landslide disaster caused a tremendous impact
58
M. Cardinali et al.: Landslide risk assessment
nationwide, which included unprecedented coverage by the
mass media. It prompted new legislation on landslide riskassessment procedures (Gazzetta Ufficiale della Repubblica
Italiana, 1998). Under the aegis of this new legislation we
have completed research aimed at assessing landslide hazard
and risk in urban and rural areas of the Umbria Region.
2
General setting
The Region of Umbria covers 8456 km2 in central Italy
(Fig. 1). The landscape is hilly or mountainous, with open
valleys and intra-mountain basins, and elevation ranging
from 50 to 2436 m a.s.l. The Tiber River, a tributary of the
Tyrrhenian Sea, drains the area. Climate is Mediterranean,
with distinct wet and dry seasons. Rainfall mainly occurs
from October to February, with cumulative annual values
ranging between 700 and 2000 mm.
In Umbria four major groups of rock units crop out,
namely: carbonate, flysch, volcanic rocks, and post-orogenic
sediments (Fig. 1a). Each lithological complex comprises
different rock types varying in strength from hard to weak
and soft rocks. Hard rocks include layered and massive
limestones, cherty limestones, sandstones, pyroclastic deposits, travertines and conglomerates. Weak rocks include
marls, shales, sands, silty clays, and overconsolidated clays.
Soft rocks are marine and continental clays, silty clays, and
shales. Rocks are mostly layered, and structurally complex
(Servizio Geologico d’Italia, 1980; Guzzetti et al., 1996;
Cardinali et al., 2001).
In Umbria, unstable slopes were recognised and studied
in many cities and towns, including Perugia, Orvieto, Assisi,
Todi, Montone, and Allerona (Felicioni et al., 1994). Geomorphological investigations revealed that landslides cover
about 14% of the entire land area. Landslide abundance and
pattern vary largely within each lithological complex that is
characterised by a prevalent geomorphological setting and by
typical geotechnical and hydrogeological properties. Failures are largely controlled by the relative position of sedimentary and tectonic discontinuities, by the relative abundance of hard versus weak or soft rocks, and by the presence
and attitude of permeable and impermeable layers (Guzzetti
et al., 1996).
3
Hazard and risk assessment project
Following the catastrophic landslide disaster in Campania
Region, on August 1998, the Italian Government passed new
legislation on landslide and flood risk assessment and mitigation (Gazzetta Ufficiale della Repubblica Italiana, 1998).
This requires that the Regional Governments and National
River Basin Authorities identify and map areas where landslide risk is most severe, and take action to reduce economic
damage and societal risk. The law was accompanied by a
“technical document” providing a general framework and
guidelines for the assessment of landslide hazard and risk
(Gazzetta Ufficiale della Repubblica Italiana, 1999).
The Government of Umbria commissioned the CNR-IRPI
Institute in Perugia to assess landslide hazard and risk in
the Region. Due to the working scale of the project (set at
1:10 000), and to economic and time constrains, not all the
regional territory could be investigated fully. The Regional
Geological Survey selected 79 places (towns, villages, individual houses, and road sections) to be investigated. The
selection was guided by pre-existing information on slope
failures, and on landslide events that caused damage. The
exact location or the extent of the areas to be investigated
was not given to us. For each site, we identified the extent of
the study area, selected the landslides for which hazard was
ascertained, and identified the vulnerable elements for which
risk had to be estimated.
4
Hazard and risk assessment strategy
Assessing landslide hazard and risk is a complex operation
that requires the combination of different techniques and
methodologies, and the interplay of various expertises, not
all of which pertain to the realm of the earth sciences (Hungr,
1997). A review of the vast literature on landslide hazard assessment (see: Varnes and IAEG, 1984; Hutchinson,
1995; Guzzetti et al., 1999; and references herein), and of
the smaller literature on landslide risk assessment (Einstein,
1988, 1997; Fell, 1994, 2000; Cruden and Fell, 1997) is not
within our scope. In this section we briefly review the concepts and terminology related to landslide hazard and risk assessment, and we present a new geomorphological methodology to evaluate landslide hazard and to estimate landslide
risk. For explanation purposes we use the village of Rotecastello in the San Venanzo Municipality, one of the 79 Umbrian localities where landslide hazard and risk were ascertained (Figs. 1b and 2).
5
Concepts and terminology
Landslide hazard refers to the natural conditions of an area
potentially subject to slope movements. It is defined as the
probability of occurrence of a landslide of a given magnitude,
in a pre-defined period of time, and in a given area (Varnes
and IAEG, 1984). The definition incorporates the concepts
of spatial location (“where”), magnitude or intensity (“how
large”), and frequency of occurrence (“when”, or “how often”). Location refers to the ability to forecast where a landslide will occur; magnitude refers to the prediction of the size
and velocity of the landslide; and frequency refers to the ability to forecast the temporal recurrence of the landslide event
(Guzzetti et al., 1999). Ideally, a landslide hazard map will
portray the location and probability of occurrence of mass
movements of pre-defined or design magnitudes in the study
area (Carrara et al., 1995, 1999; Guzzetti et al., 1999).
Landslide risk expresses the economic and social dimension of slope failure. It is generally considered to be equal to
the likelihood of death or injury, or to the expected monetary
loss due to the occurrence of a landslide (Varnes and IAEG,
M. Cardinali et al.: Landslide risk assessment
59
Fig. 1. Umbria Region. Index map and location of the study area. (a) Rock units appearing in the Umbria Region. Map Legend: 1) Postorogenic marine and continental sediments. 2) Volcanic rocks. 3) Flysch deposits; 4) Limestone and marl sediments. (b) Location of the 79
towns, villages and single dwellings where landslide hazard and risk were identified. The asterisk shows the location of Rotecastello village,
in San Venanzo Municipality, Terni Province.
1984; Einstein, 1988, 1997; Michael-Leiba et al., 1999; Fell,
1994, 2000; Fell and Hartford, 1997). Landside risk is usually defined as the product of landslide hazard and vulnerability. The latter ranges from 0, meaning no damage, to 1,
representing complete destruction. The definition of landslide risk requires that both hazard and vulnerability be defined as independent probabilities (of occurrence, for hazard;
and of damage, for vulnerability). In practice, it is rarely
possible to define hazard and vulnerability as probabilities,
which reduces the limits of rigour for applying the definition
of landslide risk.
Due to the highly variable destructiveness of landslides,
and their extremely changeable characteristics (i.e. size,
shape, velocity and momentum), vulnerability is difficult to
define precisely. Ideally, a vulnerability assessment should
include considerations of the type of failure (e.g. size, shape,
volume and velocity), the elements at risk (e.g. type, size,
construction characteristics and maintenance status), and of
a building’s ability to survive the expected landslide. This is
not an easy task, as it requires the study of past landslides that
have caused damage, as well as needing critical interpretation. The evaluation of the vulnerability of an element at risk
is further complicated by the fact that the same element may
respond well to a certain type of failure, and perform poorly
in a different type of landslide. As an example, a road may be
slightly damaged by a rock fall but completely destroyed by a
slow moving, deep-seated slump. To complicate matters further, a person travelling along the same road exhibits a high
vulnerability to rock falls (the person is likely to be killed if
hit by a rock), but has a comparatively low vulnerability to
the slow-moving slump.
6
The methodology
We assessed landslide hazard and risk using a geomorphological approach, combined with the analysis of site-specific
and historical information where this was available. We
started with the careful scrutiny and mapping of the “state
of nature”, i.e. of all the existing and past landslides that
could be identified in the study area. Based on the observed
changes in the distribution and pattern of landslides, we inferred the possible evolution of slopes, the probable shortterm types of failure and their expected frequency of occurrence. We used this information to estimate the landslide
hazard and to evaluate the landslide risk.
More precisely, the methodology involved the following
steps:
– Definition of the extent of the study area;
– Production of a multi-temporal landslide inventory map,
including a landslide classification;
– Definition of landslide hazard zones around existing
single or multiple landslides;
– Landslide hazard assessment;
60
M. Cardinali et al.: Landslide risk assessment
Fig. 2. General view of the western side
of the village of Rotecastello, in central
Umbria.
– Identification and mapping of the elements at risk, and
assessment of their vulnerability to different landslide
types;
– Evaluation of landslide risk.
7 The study area
First, the study areas had to be defined. This involved identifying the location and extent of each area to be investigated
(i.e. the “site”). This was not a trivial problem, because sites
Figure 2
in different lithological and morphological
domains had to
be identified using the same criteria.
We defined a “site” as an area bounded by drainage channels or interfluves around one of the places selected for investigation by the Regional Government. A site is an ensemble of one or more adjacent- watersheds
or “elementary
1slopes”, i.e. areas bounded by channels and interfluves.
Wherever possible major divides and drainage lines were selected. Where this was not feasible minor divides or drainage
channels were used. Selection of a geomorphological unit as
the study area is justified by the observation that landslides
occur mostly along the slope crests in the Umbria Region.
Mapping of elementary slopes or sub-watersheds was accomplished at 1:10 000 scale, using the available large-scale
topographic maps (CTR, 1:10 000 scale), locally aided by
the analysis of large- and medium-scale aerial photographs
(see Table 1). At each site, the number and the extent of
the elementary slopes depended on the local geological and
morphological setting, and on the type, number and extent
of landslides. The number of elementary slopes ranged from
one to seven, and their area ranged from 0.1 to 4 km2 .
8 Multi-temporal landslide inventory map
In Umbria landslides show a remarkable spatial recurrence
(Cardinali et al., 2000). Landslides tend to occur (in time and
in space), within or in the vicinity of other landslides, or in
the same slope or watershed. This suggests that knowledge
of the location of past failures is the key to forecasting the
future occurrence of landslides in the Region.
Within each study area we ascertained the spatial distribution of landslides through the interpretation of multiple sets
of stereoscopic aerial photographs, and by carrying out detailed field surveys in the years 2000 and 2001. Table 1 lists
the characteristics of the aerial photographs that were used.
Ten sets of photographs taken in different years were available for the period 1941–1999. The nominal scale of these
photographs ranged from 1:13 000 (Regione Umbria, 1977)
to 1:73 000 (Volo Italia, 1994). Only the GAI (1954-1955),
the Regione Umbria (1977) and the Volo Italia (1994) flights
cover the entire area. For the Rotecastello site flights A, B,
C, and H of Table 1 were used.
We started by identifying landslides on the 1954–1955
aerial photographs. Besides covering the entire region at a
scale (1:33 000) suitable for the identification of landslides,
this flight took place before the intensive post-war agricultural exploitation of the area. In the photographs landslides
are clearly identifiable because they were not obliterated by
ploughing or other farming activities (Guzzetti and Cardinali,
1989). For the Rotecastello site, Figure 3a shows the inventory map prepared by interpreting the September 1954 aerial
photographs at the 1:33 000 scale. The map portrays all the
landslides that were identified, regardless of the estimated
age. Very old (relict) deep-seated landslides are shown together with old and recent (in 1954) surficial and deep-seated
landslides. For the deep-seated landslides the crown area and
the deposits were mapped separately.
We then analysed the other sets of aerial photographs, separately and in conjunction with the 1954–1955 flight, and
we prepared separate landslide inventory maps (Figs. 3ae). The map of Fig. 3b portrays the landslides that looked
“fresh” (i.e. new or active) in the 1941 photographs. These
landslides were probably triggered by a large meteorological event that occurred in November 1937 in the Tiber basin.
Figure 3c shows all new or active landslides identified in
the 1977 photographs. These landslides were not shown on
the older photographs. An attempt (not shown in Fig. 3c)
M. Cardinali et al.: Landslide risk assessment
61
A
B
C
D
Boundary of the study area
Surficial landslide
Deep seated landslide deposit
r r
r
Crown area of deep seated landslide
Very old deep seated landslide
Crown of very old deep seated landslide
r
0
Escarpment
Rock fall area
200
400
600
800
1000 m
E
Fig. 3. Rotecastello, central Umbria. Landslide inventory maps. (a) Landslides identified on the September 1954 aerial photographs. (b)
Active and new landslides identified on the 1941 aerial photographs. (c) Active and new landslides identified on the 1977 aerial photographs.
(d) Active and new landslides identified on the 1997a aerial photographs. (e) Active and new landslides identified during the fieldwork
completed in 2000 and 2001. Rock falls, of unknown age, were identified in the field.
was made to distinguish landslides that were “fresh” in 1977
(i.e. active landslides or failures that had occurred a few
months or years before the photographs were taken) from
the other, slightly older landslides that occurred after 1954.
Figure 3d shows new landslides mapped on the 1997 aerial
photographs at the 1:20 000 scale, and triggered by the rapid
melting of snow cover in January 1997 (Cardinali et al.,
2000). Lastly, Fig. 3e shows new and active slope failures
mapped during the field campaign of 2000–2001. These
landslides occurred in the previous autumn and winter seasons. Figure 3e also shows the location of rock falls identi-
fied in the field, which are of an unknown age.
Landslide information collected through the interpretation
of aerial photographs or mapped in the field was transferred
on large-scale topographic maps (at 1:10 000 scale) or, where
these were not available, on ortophoto maps at the same
scale. The five landslide maps (Fig. 3) were then combined
to obtain a multi-temporal landslide inventory map, as shown
in Fig. 4. This was accomplished by overlaying the separate
maps, and merging them into a single map. The process required some adjustments to eliminate minor positional and
drafting errors. The multi-temporal landslide inventory map
62
M. Cardinali et al.: Landslide risk assessment
Table 1. Characteristics of the stereoscopic aerial photographs used to prepare the landslide inventory and multi-temporal landslide maps
Flight name
Scale
Type
Year
Season
Coverage
Notes
A
IGM
1:20 000
B&W
1941
–
Portions of the region
B
GAI, IGM
1:33 000
B&W
1954–1955
Various
Entire region
Flown during the
Second World War
September 1954 in
the Rotecastello area
C
D
E
F
G
H
Regione Umbria
IGM
IGM
IGM
Volo Italia
IRPI
1:13 000
1:36 000
1:36 000
1:36 000
1:73 000
1:20 000
Colour
B&W
B&W
B&W
B&W
B&W
1977
1991
1993
1994 a
1994 b
1997 a
Spring
September
July
March
September
Spring
Entire region
Northern Umbria
South western Umbria
South eastern Umbria
Entire region
Central Umbria
I
DPC
1:13 000
B&W
1997 b
Fall–Winter
Eastern Umbria
J
Regione Umbria
1:28 000
B&W
1999
October
Southern Umbria
High altitude flight
Flown after the snow-melting
event of January 1997
Flown after the
September 1997 earthquakes
Boundary of the study area
r
Escarpment
Rock fall area
2000-2001
1997
1977
1954
1941
Very old (relict) landslide
r
r
r
0
100
200
300
400
500 m
Fig. 4. Rotecastello, central Umbria. Multi-temporal landslide inventory map. Numbers in the legend refer to the year of the flights used to
identify the landslides (see Table 1).
was then digitised and stored in a GIS for analysis and display.
9 Landslide classification
Landslides were classified according to their type of movement, and their estimated ages, degrees of activity, depths,
and velocities. The level of certainty in the recognition of
the landslide was also noted. Landslide type was defined according to the classifications of Varnes (1978), Cruden and
Varnes (1996), and the WP/WLI (1990, 1993, 1995). Landslide age, activity, depth, and velocity were ascertained according to the type of movement, the morphological characteristics and appearance of the landslide, the local lithological and structural settings, and, where available, the results
M. Cardinali et al.: Landslide risk assessment
of site-specific investigations carried out to solve local instability problems.
The relative age of a landslide was defined as recent, old
or very old. Landslides of recent age were recognised according to the following categories: morphologically fresh
(i.e. “active”) on the aerial photographs taken in 1941 and
1954–1955; new landslides or reactivations of pre-existing
landslides in the more recent aerial photographs (1977, 1991,
1993, 1997); or by field mapping carried out in 2000 and
2001. Old landslides were recognised on the 1954–1955
aerial photographs, but these features did not exhibit any detectable morphological change of the (entire) original feature in the more recent aerial photographs or during the field
surveys. Very old landslides were identified on the 1954–
1955 aerial photographs, and were regarded as those that
showed a relict morphology largely dismantled by erosion.
They mostly involved large volumes of rock or sediment, and
probably occurred in different climatic or seismic conditions
(Guzzetti et al., 1996).
The evaluation of landslide activity requires detailed information on the movement of the landslide (WP/WLI, 1993;
1995). This implies inevitably that some device or system
capable of measuring displacement (e.g. inclinometer, extensiometer or topographic survey) is available. Given the scale
of the investigation, the extent and number of landslides in
each study area, and the lack of quantitative measurements,
landslide activity was ascertained mostly on a geomorphological base; i.e. landslides were classified as active where
they looked fresh on the aerial photographs, or where movement was known from monitoring systems.
Landslides were classified as deep-seated or shallow, depending on the type of movement and the landslide volume.
Evaluation of landslide volume was based on the type of failure, and the morphology and geometry of the detachment
area and deposition zone.
Landslide velocity was considered as a proxy for landslide
type, and was classified accordingly (WP/WLI, 1995). Rotational or translative slides, earth-flow slides, flows, and complex or compound slides were classified as slow moving failures. Debris flows were classified as rapid movements. Rock
falls, including topples, were classified as fast moving landslides.
We acknowledge that landslide classification, and in particular, the evaluation of landslide age, activity, velocity and
depth, includes simplifications. The classification required
geomorphological inference on our side, but it does fit the
available information on landslide types and process in Umbria (Felicioni et al., 1994; Guzzetti et al., 1996; Alexander,
2000).
10
Landslide frequency
To assess hazard, information on landslide frequency is
needed. Frequency refers to the temporal occurrence of landslides and can be obtained through the analysis of historical
data (Guzzetti et al., 1999). In general, a complete (or at
63
least systematic) record of past landslides from which to derive their frequency of occurrence is difficult to obtain for a
single landslide, slope, or small watershed (Ibsen and Brunsden, 1996; Glade, 1998; Guzzetti et al., 1999). Guzzetti et
al. (1999) argued that evidence of past movements on a slope
might not necessarily indicate the possibility of future landslides.
Due to a lack of information on the temporal occurrence
of landslides for most of the investigated sites, we ascertained landslide frequency based on the analysis of the multitemporal inventory map, which covers an observation period
of 60 years. We ascertained the frequency of occurrence of
single or multiple landslides, based on the number of events
(i.e. the localized morphological changes) recognised during this observation period. For convenience, four classes of
landslide frequency were identified:
– Low frequency (1), when only one landslide event was
observed;
– Medium frequency (2), when 2 events were observed;
– High frequency (3), when 3 events were observed; and
– Very high frequency (4), when more than 3 events were
observed in 60 years.
No distinction was made between the date of occurrence
of slope failures inferred through the interpretation of aerial
photographs or known from field instrumentation or technical reports. A site where a landslide was only recognised in
1954–1955 and one where a landslide was identified only on
the 1997 aerial photographs were both assigned a low frequency. Similarly, a site where landslides were observed
in 1941, 1954–1955, 1977 and in 2001 (i.e. 4 times in 60
years), and a site where landslides were identified in 1977,
1994, 1997, and 2000 (i.e. 4 times in 24 years), were assigned the same (very high) frequency. A further complication or added uncertainty arose because the landslide dates
inferred from the multi-temporal inventory map had different temporal constrains. A landslide identified on the aerial
photographs taken immediately after an event (i.e. 1997a,
1997b), or not too far after it (i.e. 1941), provided a closer
estimate on the date of movement than a landslide identified
on the other flights, taken decades (or even centuries for the
very old landslides) after the slope failures.
11
Landslide intensity
The definition of hazard requires information on landslide
intensity. Contrarily to other natural hazards, such as earthquakes or volcanic eruptions, no unique or commonly recognised measure of landslide intensity is available (Hungr,
1997). Since our goal was to estimate landslide risk, we considered landslide intensity (I ) as a measure of the destructiveness of the landslide (Hungr, 1997), and we defined it as
a function of the landslide volume (v) and of the landslide
64
M. Cardinali et al.: Landslide risk assessment
Boundary of the study area
Slight
Medium
High
Very high
#
#
Escarpment
High (for rock fall)
#
#
0
100
200
300
400
500 m
Fig. 5. Rotecastello, central Umbria. Map showing the expected landslide intensity (see Table 2). Landslides were also classified on the
basis of estimated age, degree of activity, depth of movement and velocity (not shown here). Rock falls, shown by asterisks, were assigned
high landslide intensities.
Table 2. Landslide intensity, grouped into four classes: light (1), medium (2), high (3) and very high (4). Note that landslide intensity varies
with the landslide type
Estimated volume
(m3 )
< 0.001
< 0.5
> 0.5
< 500
500–10 000
10 000–50 000
> 500 000
>> 500 000
Fast moving landslide
(Rock fall)
Slight (1)
Medium (2)
High (3)
High (3)
High (3)
Very High (4)
Expected landslide velocity
Rapid moving landslide Slow moving landslide
(Debris flow)
(Slide)
expected velocity (s),
I = f (v, s).
Table 2 shows how we assigned the intensity to each landslide (or group of landslides) based on the estimated volume
and the expected velocity. We estimated volume on the basis
of landslide type. For slow-moving slides, volume depended
on the estimated depth of movements; for rapid moving debris flows it depended on the size of the catchment and the es-
Slight (1)
Medium (2)
High (3)
Very High (4)
Slight (1)
Medium (2)
High (3)
Very High (4)
timated volume of debris in source areas and along channels;
while for fast-moving rock falls it depended on the maximum
size of a single block as estimated from field observations.
The expected landslide velocity depends on the type of failure, its volume and the estimated depth of movement. For a
given landslide volume, fast moving rock falls have the highest landslide intensity, while rapidly moving debris flows exhibit intermediate intensity, and slow moving landslides have
the lowest intensity.
M. Cardinali et al.: Landslide risk assessment
Figure 5 shows the distribution of expected landslide intensity for the Rotecastello site.
12
65
Table 3. Landslide hazard for each LHZ: Landslide intensity,
grouped into four classes: light (1), medium (2), high (3) and very
high (4), and the estimated landslide frequency, grouped into four
classes: low (1), medium (2), high (3) and very high (4)
Landslide hazard zones
The evaluation of landslide hazard can be carried out regionally or locally. The first involves assessing landslide hazard
for an entire area, such as a river basin, a municipality or a
province; the latter involves assessing the hazard of a single landslide, or a group of landslides (Carrara et al., 1995,
1999; Hutchinson, 1995; Guzzetti et al., 1999). Making a
compromise, we decided to evaluate landslide hazard only in
the areas of evolution of existing (mapped) landslides. For
this purpose, a “landslide hazard zone” (LHZ) is defined as
the area of possible (or probable) short-term evolution of an
existing landslide, or a group of landslides, of similar characteristics (i.e. of type, volume, depth, and velocity), identified
from the aerial photographs or observed in the field. A LHZ
is therefore a “landslide scenario”, delimited using geomorphological criteria.
To identify and map the extent of LHZs we used the multitemporal landslide inventory map. Based on the observed location, distribution and pattern of landslides, their observed
or inferred style of movement and activity, and the local
lithological and morphological setting, we mapped the area
of possible evolution of each landslide, or group of landslides
within each elementary slope. To define the LHZs, we considered the observed partial or total reactivation of existing
landslides; the lateral, head (retrogressive) or toe (progressive) expansion of the existing landslides; and the possible
occurrence of new landslides of similar type and intensity.
We identified different landslide scenarios (i.e. different
LHZ) for each type of failure observed on an elementary
slope (e.g. fast-moving rock falls, rapid-moving debris flows,
slow-moving earth-flow slumps or compound failures). An
LHZ includes the area currently recognised as a landslide
(i.e. the crown and the deposit), and the area of possible direct or indirect influence of the observed phenomenon. For
slow-moving failures, such as slides, slumps, earth-flows or
compound failures, the LHZ is generally limited to the immediate surroundings of the existing landslide, or group of landslides (cf. Fig. 6c). This is justified by the fact that in Umbria
the evolution of these types of landslides is relatively slow
and spatially predictable (Cardinali et al., 2000). For relict
(i.e. very old) landslides, most probably triggered in different morphological, climatic or seismic conditions (Guzzetti
et al., 1996), the LHZ coincided with the entire slope element (cf. Fig. 6a). For debris flows the LHZ encompassed
the source areas, the river channels and the depositional areas
(alluvial or debris fans). For rock falls, topples, and minor
rockslides, the LHZ included the escarpments from where
landslides detached, and the talus, debris cones, and debris
slopes along which rock falls travelled and where they eventually stopped (cf. Fig. 6d).
Estimated landslide
frequency
Low (1)
Medium (2)
High (3)
Very high (4)
13
Light
(1)
11
21
31
41
Landslide intensity
Medium High Very high
(2)
(3)
(4)
12
13
14
22
23
24
32
33
34
42
43
44
Hazard assessment
Landslide hazard (H ) depends on the frequency of landslide
movements (F ) and on the landslide’s intensity (I ),
H = f (F, I ).
Table 3 shows how we defined landslide hazard for each
LHZ, combining frequency and intensity. Landslide frequency was estimated using four classes, based on the number of landslide events (of the same type) observed within
each LHZ. Landslide intensity was defined in four classes,
based on the estimated volume and the expected velocity.
Figure 6 shows the hazard at the Rotecastello site for the
different types of failures: very old (relict), deep-seated
landslides (Fig. 6a); old and recent, deep-seated landslides
(Fig. 6b); shallow, mostly recent, landslides (Fig. 6c); and
rock falls (Fig. 6d).
Levels of landslide hazard in Table 3 are shown using a
two-digit positional index. The right digit shows the landslide intensity (I ) and the left digit shows the estimated landslide frequency (F ). The index expresses landslide hazard by
keeping the two components of the hazard distinct from one
another. This facilitates landslide hazard zoning by allowing
a user to understand whether the hazard is due to a high frequency of landslides (i.e. high recurrence), a large intensity
(i.e. large volume and high velocity), or both.
It is worth noting that values of the landslide hazard index shown in Table 3 do not provide an absolute ranking of
hazard levels. If the extreme values are easily defined, intermediate conditions of landslide hazard are more difficult to
rank. A landslide that exhibits low frequency and light intensity (H = 11) will certainly have a much lower hazard than
one that exhibits very high frequency and intensity (H = 44).
Deciding whether the hazard of a landslide with very high
frequency and light intensity (H = 41) is higher (or lower)
than that of a landslide with low frequency and very high intensity (H = 14) is not straightforward and may be a matter
of opinion.
14
Elements at risk and their vulnerability
We had no access to any map of the elements at risk (e.g.
houses, buildings, roads, railways, utilities and population)
66
M. Cardinali et al.: Landslide risk assessment
A
B
13
12
13
14
r r
r
12
r r
r
13
12
13
0
200
400
600 m
0
200
400
600 m
C
D
21
22
12
21
12
r r
r
r r 33
r
22
41
12
12
11
21
12
11
11
12
0
200
400
600 m
0
200
400
600 m
Fig. 6. Rotecastello, central Umbria. Maps of landslide hazard zones (LHZ, grey pattern) and of landslide hazard. (a) LHZ for very old
(relict) slow moving, deep-seated landslide. (b) LHZ for old and recent, slow moving, deep-seated slides. (c) LHZ for slow-moving, surficial
landslides. (d) LHZ for fast moving rock falls. Landslide hazard is shown using a two-digit index. Where two values are given (e.g. (b) or
(c)) landslides of the same type but of different intensity or frequency are present.
Table 4. Types of element at risk (for structures and infrastructures)
Code
Frequency
HD
LD
Built-up areas with high population density
Built-up areas with low population density
and scattered houses
Industries
Animal farms
Sports facilities
Quarries
Main roads, motorways, highways
Secondary roads
Farm and minor roads
Railway lines
Cemeteries
IN
FA
SP
Q
MR
SR
FR
RW
C
at an appropriate scale and with sufficient detail and accuracy. For each study area we prepared a map of the elements
at risk at the 1:10 000 scale based on information on built-up
areas, structures and infrastructures present on large-scale topographic maps (CTR at 1:10 000 scale), aided by the recent
aerial photographs (Fig. 7). Care was taken in locating the
elements at risk inside or in the vicinity of the landslides or
in potentially dangerous areas. The map was then digitised
and stored in a GIS for analysis and display.
Elements at risk were classified according to a simple legend, as shown in Table 4. Of the eleven classes present in
the legend, six referred to built-up areas and structures (i.e.
houses, buildings, industry and farms, sport utility, cemetery); four to transportation utilities (roads and railways); and
one to mining activities (a quarry). Built-up areas included
public streets, roads, squares and gardens.
No information was available to us on the amount (or density) of the population in the study areas. To estimate the
landslide risk to the population we considered the houses,
buildings and roads as a proxy for the population density. In
other words, we considered the population to be vulnerable
in conjunction with (or because of) the presence of structures
and infrastructures. As an example, vulnerability of the population was considered higher along a high transit road than
along a secondary road. For sparse, farming structures, it was
considered lower than in a densely populated zone.
As previously discussed, evaluating the expected damage
to each element at risk is a difficult and uncertain operation.
M. Cardinali et al.: Landslide risk assessment
67
HD
LD
MR
SR
FR
A
D
B
C
B
0
100
200
300
400
500 m
Fig. 7. Rotecastello, central Umbria. Map of the elements at risk. Legend (see Table 4): HD, built-up areas with a high population density;
LD, built-up areas with a low population density and scattered houses; MR, main road; SR, secondary road; FR, farm road. Letters indicate
the location of elements at risk in Figs. 8a-d.
Table 5. Vulnerability, the expected damage to the elements at risk. A = superficial (aesthetical, minor) damage; F = functional (or medium)
damage; S = structural (or total) damage. For classes of elements at risk see Table 4. For landslide intensity see Table 2
Landslide Intensity
Light
Medium
High
Very high
Rock fall
Debris flow
Slide
Rock fall
Debris flow
Slide
Rock fall
Debris flow
Slide
Rock fall
Debris flow
Slide
HD
A
A
A
F
F
F
S
S
S
S
S
S
Elements at Risk
Structures and infrastructures
Buildings
Roads
LD IN FA SP C MR SR FR RW
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
F
F
A
A
A
A
A
A
A
F
S
A
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
S
S
F
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
To estimate vulnerability we used a simple approach based
on the inferred relationship between the intensity and type of
the expected landslide, and the likely damage the landslide
would cause. Table 5 shows the expected damage to buildings and roads, and to the population if (i.e. where) affected
by landslides of different type (rock fall, debris flow or slide)
and intensity (slight, medium, high, or very high). The latter
Population
Others
Q
A
A
A
F
F
F
S
S
S
S
S
S
Direct
No
No
No
Yes
Yes
No
Yes
Yes
No
Yes
Yes
No
Indirect
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Homeless
No
No
No
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
is a function of landslide volume and velocity. Table 5 was
prepared based on the review of the limited literature on the
subject (Alexander, 1989; Michael-Leiba et al., 1999; Fell,
2000), the analysis of the damage caused by slope failures
in Umbria (Felicioni et al., 1994, Alexander, 2000; Cardinali
et al., 2000; Antonini et al., 2001), and our experience and
judgement. A crude estimate (i.e. few, many, very many) of
68
M. Cardinali et al.: Landslide risk assessment
the number of people potentially subject to landslide risk was
also considered.
Damage to structures and infrastructures was classified as:
– Superficial (aesthetic, minor) damage, where the functionality of buildings and roads is not compromised, and
the damage can be repaired, rapidly and at low cost;
– Functional (medium) damage, where the functionality
of structures or infrastructures is compromised, and the
damage takes time and large resources to be fixed;
– Structural (severe or total) damage, where buildings or
transportation routes are severely or completely damaged, and require extensive (and costly) work to be
fixed. In this category, demolition and reconstruction
may be required.
According to Table 5 a rock fall of light intensity is capable
of causing superficial (minor) damage to buildings and roads;
and a rock fall of high intensity (i.e. of larger volume) causes
severe structural damage. A shallow slide of medium intensity causes aesthetic damage to buildings, functional damage to industrial and farming activities and to sport facilities,
functional damage to major roads, and structural damage to
other roads and railways. A deep-seated slide of very high
intensity causes structural (i.e. severe) damage to all types of
structures and infrastructures.
The expected damage to the population was classified as:
– Direct, where casualties are expected;
– Indirect, where only socio-economic damage is expected; and
– Temporary or permanent loss of private houses resulting
in evacuees and homeless.
Direct damage to the population is foreseen for rapid and
fast moving landslides, or for high intensity, slow moving
ones. Indirect damage to the population is expected where
landslides can cause functional or structural damage to the
infrastructure, with negative socio-economic effects upon
public facilities. Homelessness is expected where functional
or structural damage to buildings is foreseen.
15
Risk evaluation
As previously discussed, a rigorous assessment of landslide
risk is difficult to achieve. We argue that even where hazard
and vulnerability cannot be determined as (numerical) probabilities, landslide risk still depends on the “state of nature”
(i.e. landslide hazard, H ) and on the expected damage (i.e.
the vulnerability, V ), or
Rs = f (H, V ).
We used this more general relationship to ascertain the
specific landslide risk (Rs ), i.e. the risk to which a set of
elements (e.g. building, roads, etc.) is subject when a landslide occurs (Einstein, 1988). We defined the specific risk
separately for each class of elements at risk (Table 4), within
each LHZ (Fig. 8).
To help assign a value of specific landslide risk to each
element, we prepared Table 6. The table correlates the expected damage (i.e. aesthetic, functional or structural) to the
landslide hazard. The latter was loosely ranked from low
(11) to high (44) values. Construction of Table 6 required
extensive discussion. It is largely based on the analysis of
damage caused by two recent regional landslide events: a
rapid snow melt that triggered thousands of failures in January 1997 (Cardinali et al., 2000), and the Umbria-Marche
earthquakes of September–October 1997 that mostly caused
rock falls (Antonini et al., 2001). Information on past landslide damage in the Umbria region was also considered (Felicioni et al., 1994; Alexander, 2000).
To show the level of specific risk, we added to the left of
the two-digit landslide hazard index a third (alphanumerical)
digit describing the expected damage (i.e. aesthetic, functional or structural, see Table 5). Thus, the specific risk index shows, from right to left, the landslide frequency, the
landslide intensity, and the expected damage caused by the
specific type of landslide (Table 6 and Fig. 8). If more than
a single class of elements at risk is present in a LHZ, a different value of specific risk is computed for each class (e.g.
Fig. 8b).
Figure 8 shows examples of specific landslide risk for various vulnerable elements in the Rotecastello area. Figure 8a
shows that landslide risk for a major road (MR) is S14, because the expected damage is structural (S) and the hazard
index is 14 (i.e. frequency is low and intensity is very high).
In Fig. 8b landslide risk to a secondary road (SR) is defined
as S12 because the expected damage is structural (S) and the
hazard index is 12 (i.e. low frequency and medium intensity).
In the same LHZ, landslide risk for the low-density settlement of Podere Piano (LD) is ascertained as A12, because
the expected damage is superficial (A) and the hazard index
is 12. Figure 8c shows that landslide risk to a secondary road
(SR) is F11, because the expected damage is functional (F)
and the hazard index is low (i.e. 11 – low frequency and
slight intensity). Lastly, Fig. 8d shows that landslide risk to
the high-density settlement of Rotecastello (HD) is S33, because the expected damage is structural (S) and the hazard
index is 33, since both the landslide frequency and the landslide intensity are high.
As for the landslide hazard index, the landslide risk index (Rs ) does not provide an absolute ranking of risk levels.
The “extreme” conditions are easily ranked: a house having
an Rs of A11 (i.e. aesthetic damage due to a low frequency
and slight intensity slope failure) has a lower specific landslide risk than a dwelling with Rs = S44 (i.e. structural damage caused by a very high frequency and very high intensity
landslide). It is not easy to decide for the several intermediate conditions. Decisions should be made on a case-by-case
basis, considering the type of elements at risk, their vulner-
M. Cardinali et al.: Landslide risk assessment
A
69
B
SR, Rs = S12
MR, Rs = S14
H = 12
H = 14
0
LD, Rs = A12
0
100 m
C
100 m
D
SR, Rs = F11
HD, Rs = S33
H = 33
H = 11
0
100 m
0
100 m
Fig. 8. Rotecastello, central Umbria. Maps of specific landslide risk (Rs). Grey pattern indicates the extent of the LHZ. (a) Landslide risk to
a major road (MR). (b) Landslide risk to a secondary road (SR) and to a low-density settlement (LD). (c) Landslide risk to a secondary road
(SR). (d) Landslide risk to the high-density settlement of Rotecastello (HD).
ability, the possible mitigation measures, and the economic
and social implications of landslide risk.
Definition of specific landslide risk levels may not be
enough where economic decisions must be taken. The Landslide Risk-Assessment and Reduction Act (Gazzetta Ufficiale della Repubblica Italiana, 1998) required the ranking of
the most dangerous landslide areas according to the expected
(total) landslide risk. Although we do not think that this is a
wise way of dealing with landslide risk, we devised a system
to aggregate the detailed information given by the specific
landslide risk index into one of the four classes of landslide
risk required by the law. Very high landslide risk (severe risk,
R4) was assigned where rapid and fast-moving landslides
could cause direct damage to the population. These were the
areas where debris flows and rock falls could cause casualties or homelessness. High landslide risk (R3) was assigned
to the areas where slow-moving landslides could cause structural and functional damage to structures and infrastructure.
In these areas casualties are not expected. Moderate landslide risk (R2) was attributed where aesthetic damage to vulnerable elements is expected, caused by slow-moving slope
failures by fast or rapid moving landslides of slight intensity. Lastly, low landslide risk (R1) was assigned to the areas
where no element at risk is currently present within a landslide hazard zone.
16
Discussion
At the 79 study areas considered in this work we identified
980 landslide hazard zones (LHZ), pertaining to all the lithological complexes and major landform types present in Umbria (Fig. 1). Due to funding and time constraints, landslide
risk was ascertained for a subset of 210 LHZ (21.4%), covering a total of about 20 km2 . The data set is large enough to
enable some general conclusions to be drawn.
70
M. Cardinali et al.: Landslide risk assessment
Table 6. Levels of specific landslide risk, based on landslide hazard, in 16 classes (see Table 3), and vulnerability, in 3 classes (see Table 5).
Classes of specific risk assigned to the 79 study areas are shown in bold type
Hazard
low
↑
↓
high
11
12
13
21
14
22
23
31
32
24
33
41
42
34
43
44
Vulnerability (expected damage)
(Minor) damage (Major) damage (Total) damage
A11
F11
S11
A12
F12
S12
A13
F13
S13
A21
F21
S21
A14
F14
S14
A22
F22
S22
A23
F23
S23
A31
F31
S31
A32
F32
S32
A24
F24
S24
A33
F33
S33
A41
F41
S41
A42
F42
S42
A34
F34
S34
A43
F43
S43
A44
F44
S44
The proposed geomorphological method is empirical and
subject to various levels of uncertainty, but has proved to be
reliable and cost effective, allowing for a detailed definition
of landslide hazard and risk in urban and rural areas. The
method allows for the comparison of landslide hazard and
risk in different (and distant) areas, and where different landslide types are present.
Assessment of landslide hazard requires forecasts to be
made in different settings in space and time, and with different types and dimensions. It can be carried out over a
large area, such as a drainage basin or province, or for a
single landslide or group of landslides (Cruden and Fell,
1997; Guzzetti et al., 1999). The proposed method ascertains landslide hazard in the areas of (probable) evolution
of the existing landslides, and for the various types of failures (i.e. slides, debris flows, rock falls) separately. The
method says nothing about the hazard outside a LHZ, even
within the same elementary slope. In these areas minor landslides, mostly superficial failures can occur with a low frequency. For a regional, spatially distributed landslide hazard and risk assessment, other methods should be used (van
Westen, 1994; Carrara et al., 1995, 1999; Guzzetti et al.,
1999), possibly in combination with the method proposed
here.
The methodology requires extensive geomorphological
judgment. For this reason it should only be used by skilled
geomorphologists. If the extent, type, distribution and pattern of past and present landslides are not correctly and fully
identified, serious errors can occur and thus affect the estimate of landslide hazard and risk.
With this in mind, the definition of the temporal frequency
of landslides from the analysis of the multi-temporal inventory map is particularly important. The map covers a period
of 60 years (1941–2001), which is long enough to evaluate
the short-term evolution of slopes in the areas investigated.
Where information on landslide frequency is available for a
shorter period of time (10–15 years or less), the reliability
of the hazard forecast is reduced. If a landslide event fails
to be recognized, the frequency of occurrence is underestimated, and hazard and risk estimates are negatively affected.
It should be noted that the method estimates the expected
landslide frequency based on what has happened (and was
observed) in the recent past. If low frequency, high magnitude events did not occur (or were not recognised) in a LHZ,
the hazard assessment in the area may be biased, and the actual risk underestimated. This is a limitation of the method.
Uncertainty varies with the different steps of the method.
The production of both the separate landslide inventory
maps, and the (combined) multi-temporal landslide map was
less uncertain than the identification of the landslide hazard zones, or the possible spatial evolution of the existing
landslides, which were obtained mostly through geomorphological inference. Landslides mapped through the interpretation of aerial photographs were carefully checked in the field,
whereas the identification and mapping of LHZs was based
on the observation of other landslides and on the inferred geomorphological evolution of slopes. Estimates of landslide
volume and velocity, which are essential for the evaluation
of landslide intensity, also exhibited uncertainty.
The method relies on correlation tables, which are used
to define landslide intensity (Table 2), ascertain landslide
hazard (Table 3), evaluate the expected damage to the vulnerable elements (Table 5), and estimate landslide risk (Table 6). These tables are based on empirical observations and
on our own experience, but are also the result of a heuristic
approach. Whenever possible, we tried several possibilities
and evaluated the difference after each attempt. We believe
that the tables fit the present understanding of landslide processes and match landslide damage in Umbria satisfactorily.
However, the tables should not be considered definitive
M. Cardinali et al.: Landslide risk assessment
and should not be used unconditionally in all settings. If applied to other sites, or in other study areas, the tables should
be carefully checked with the local information on landslide
types and damage. If one or more of the tables is changed
significantly, the hazard and risk assessment will vary, and
may not be comparable to the one we have prepared. This is
particularly important for Tables 5 and 6.
Landslide hazard and risk are expressed using a multipledigit index that portrays, in a compact format, all the variables used to ascertain landslide hazard and risk. The index
allows for the ranking of risk conditions at the end of the risk
assessment process, when all the necessary information is
available, and not “a priori”, based on pre-defined categories.
We consider this a major advantage of the method, giving decision makers great flexibility in deciding which area exhibits
the highest risk, and providing geologists and engineers with
a clue about why any given vulnerable element is at risk.
Lastly, the proposed method is not simple. For a dependable and consistent prediction it requires multiple sets of
aerial photographs, and a team of experienced geomorphologists to interpret them. This cannot be considered a limitation: landslide hazard and risk assessments are difficult tasks,
and require proper expertise and skills.
17
Conclusions
We have presented a geomorphological method to ascertain
landslide hazard and to evaluate the associated risk in the
Umbria Region. The method is based on careful recognition
of present and past landslides, scrutiny of the local geological and morphological settings, and analysis of site-specific
and historical information on past landslide events. Most
of the information used to ascertain landslide hazard is obtained from the careful analysis of a multi-temporal landslide inventory map that portrays information on the distribution, type and pattern of landslides, and on their changes
in time. The resulting map is obtained by merging landslide
inventory maps prepared through the analysis of stereoscopic
aerial photographs of different ages, together with field surveys.
The method was applied in 210 landslide hazard zones located around or in the vicinity of 79 towns and villages in the
Umbria Region. In these areas landslide hazard was ascertained, vulnerable elements were identified, and specific risk
was evaluated. The results prove that, although it is empirical
and subject to various levels of uncertainty, the method can
provide reasonable estimates of landslide hazards and risk in
urban and rural areas. At present it is not possible to judge
quantitatively how good the proposed method is. Reliability
and effectiveness of the maps will have to be evaluated by
Town Officials, private consultants involved in land use and
city planning, and other impartial observers. The time and
human resources required for completing the risk assessment
procedure at each site averaged 5 days for a team of 3–4 people, including bibliographical investigation, interpretation of
71
the aerial photographs, field surveys, and the production of
the final maps in digital format.
Acknowledgements. The research was supported by grants from the
Regione dell’Umbria and the Tiber River Basin Authority. We are
grateful to D. Alexander and E. Brabb for reviewing the manuscript.
This paper is CNR-GNDCI publication number 2458.
References
Alexander, D.: Urban landslides, Progress in Physical Geography,
13, 2, 157–191, 1989.
Alexander, D.: Landslide risk estimation in Umbria Region. Unpublished technical report for the CNR-IRPI, 110, 2000.
Antonini, G., Ardizzone, F., Cardinali, M., Galli, M., Guzzetti,
F., and Reichenbach, P.: Surface deposits and landslide inventory map of the area affected by the 1997 Umbria-Marche earthquakes. Bollettino Societá Geologica Italiana, in press, 2001.
Cardinali, M., Ardizzone, F., Galli, M., Guzzetti, F., and Reichenbach, P.: Landslides triggered by rapid snow melting: the December 1996-January 1997 event in Central Italy. Claps P. and
Siccardi F. (eds.), Mediterranean Storms, Proceedings Plinius
Conference’99, Maratea, 14–16 October 1999, CNR GNDCI
publication number 2012. Bios, Cosenza, 439–448, 2000.
Cardinali, M., Antonini, G., Reichenbach, P., and Guzzetti, F.:
Photo-geological and landslide inventory map of the Upper Tiber
River basin. CNR GNDCI publication number 2154, map at
1:100 000 scale, 2001.
Carrara, A., Cardinali, M., Guzzetti, F., and Reichenbach, P.: GIS
technology in mapping landslide hazard. In: Carrara, A. and
Guzzetti, F. (eds.), Geographical Information Systems in Assessing Natural Hazards. Kluwer Academic Publisher, Dordrecht,
The Netherlands, 135–175, 1995.
Carrara, A., Guzzetti, F., Cardinali, M., and Reichenbach, P.: Use of
GIS Technology in the Prediction and Monitoring of Landslide
Hazard. In: Carrara, A. and Guzzetti, F. (eds.), Natural Hazards,
special issue on Techniques and Tools for Mapping Natural Hazards and Risk Impact on the Developed Environment, 20, 2–3,
117–135, 1999.
Catenacci, V.: Il dissesto geologico e geoambientale in Italia dal
dopoguerra al 1990. Memorie Descrittive della Carta Geologica d’Italia, Servizio Geologico Nazionale, (in Italian), 47, 301,
1992.
Crescenti, U.: La grande frana di Ancona del 13 dicembre 1982.
Studi Geologici Camerti, Special Issue, Camerino, (in Italian), 1,
1986.
Cruden, D. M. and Fell, R. (eds.): Landslide risk assessment.
Proceedings International Workshop on Landslide Risk Assessment, Honolulu, 19–21 February 1997, Balkema, Rotterdam,
371, 1997.
Cruden, D. M. and Varnes, D. J.: Landslide types and processes. In:
Turner, A. K. and Schuster, R. L. (eds.), Landslides, Investigation
and Mitigation. Transportation Research Board, Special Report
247, Washington D. C., 36–75, 1996.
Einstein, H. H.: Special lecture: Landslide risk assessment procedure. Proceedings 5th International Symposium on Landslides,
Lausanne, 2, 1075–1090, 1988.
Einstein, H. H.: Landslide Risk – Systematic approaches to assessment and management. In: Cruden, D. M. and Fell, R. (eds.),
Landslide Risk Assessment. Proceedings International Work-
72
shop on Landslide Risk Assessment, Honolulu, 19–21 February
1997, Balkema, Rotterdam, 25–50, 1997.
Fell, R.: Landslide risk assessment and acceptable risk. Canadian
Geotechnical Journal. 32, 2, 261–272, 1994.
Fell, R. and Hartford, D.: Landslide risk management. In: Cruden,
D. and Fell, R. (eds.), Proceedings International Workshop on
Landslide Risk Assessment, Honolulu, 19–21 February 1997,
Balkema, Rotterdam, 51–109, 1997.
Fell, R.: Landslide risk management concepts and guidelines. Australian Geomechanics Society, Sub-committee on landslide risk
management, 69, 2000.
Felicioni, G., Martini, E., and Ribaldi, C.: Studio dei Centri Abitati
Instabili in Umbria. Atlante regionale: CNR GNDCI publication
number 979, Rubettino Publisher, (in Italian), 418, 1994.
Gazzetta Ufficiale della Repubblica Italiana: Misure urgenti per la
prevenzione del rischio idrogeologico ed a favore delle zone colpite da disastri franosi nella regione Campania. Serie Generale,
Anno 139, n. 208, 7 September 1998, (in Italian), 53–74, 1998.
Gazzetta Ufficiale della Repubblica Italiana: Atto di indirizzo e coordinamento per l’individuazione dei criteri relativi agli adempimenti di cui all’art. 1, commi 1 e 2, del decreto-legge 11 giugno
1998, n. 180. Serie Generale, Anno 140, n. 3, 5 January 1999, (in
Italian), 8–34, 1999.
Glade, T.: Establishing the frequency and magnitude of landslidetriggering rainstorm events in New Zealand. Environmental Geology, 35, 2–3, 160–174, 1998.
Guzzetti, F.: Landslide fatalities and evaluation of landslide risk in
Italy, Engineering Geology, 58, 89–107, 2000.
Guzzetti, F. and Cardinali, M.: Carta Inventario dei Fenomeni Franosi della Regione dell’Umbria ed aree limitrofe. CNR GNDCI
publication number 204, 2 sheets, scale 1:100 000, (in Italian,
1989.
Guzzetti, F., Cardinali, M., and Reichenbach, P.: The influence of
structural setting and lithology on landslide type and pattern. Environmental and Engineering Geoscience, 2, 4, 531–555, 1996.
Guzzetti, F., Carrara, A., Cardinali, M., and Reichenbach, P.: Landslide hazard evaluation: an aid to a sustainable development. Geomorphology, 31, 181–216, 1999.
Hungr, O.: Some methods of landslide hazard intensity mapping.
In Cruden, D. M. and Fell, R., (eds.), Landslide risk assessment.
M. Cardinali et al.: Landslide risk assessment
Proceedings International Workshop on Landslide Risk Assessment, Honolulu, 19–21 February 1997, Balkema, Rotterdam,
215-226, 1997.
Hutchinson, J. N.: Keynote paper: Landslide hazard assessment. In:
Bell (ed.), Landslides, Balkema, Rotterdam, 1805–1841, 1995.
Ibsen, M.L. and Brunsden, D.: The nature, use and problems of
historical archives for the temporal occurrence of landslides, with
specific reference to the south coast of Britain, Ventnor, Isle of
Wight, Geomorphology, 15, 241–258, 1996.
Michael-Leiba, M., Baynes, F., and Scott, G.: Quantitative landslides of Cains. Australian Geological Survey Organisation, Department of Industry, Science & Resources. AGSO Records
1999/36, 40, 1999.
Regione Piemonte: Eventi Alluvionali in Piemonte. 2–6 novembre
1994, 8 luglio 1996, 7–10 ottobre 1996. Direzione Servizi Tecnici di Prevenzione, Torino, (in Italian), 415, 1998.
Servizio Geologico d’Italia: Carta Geologica dell’Umbria. Map at
1:250 000 scale, (in Italian), 1980.
van Westen, C. J.: GIS in landslide hazard zonation: A review with
examples from the Colombian Andes. In: Price, M. F. and Heywood, D. I. (eds.), Taylor and Francis, London, 135–165, 1994.
Varnes, D. J.: Slope movements, type and processes. In: Schuster, R.L. and Krizek, R.J. (eds.), Landslide analysis and control.
Transportation Research Board, National Academy Sciences,
Washington, D. C., Special report 176, 11–33, 1978.
Varnes, D. J. and IAEG Commission on Landslides and other MassMovements: Landslide hazard zonation: a review of principles
and practice. UNESCO Press, Paris, 63, 1984.
WP/WLI – International Geotechnical societies’ UNESCO Working Party on World Landslide Inventory: A suggested method
for reporting a landslide. International Association Engineering
Geology Bulletin, 41, 5–12, 1990.
WP/WLI – International Geotechnical societies’ UNESCO Working Party on World Landslide Inventory: A suggested method
for describing the activity of a landslide. International Association Engineering Geology Bulletin, 47, 53–57, 1993.
WP/WLI – International Geotechnical societies’ UNESCO Working Party on World Landslide Inventory: A suggested method
for describing the rate of movement of a landslide. International
Association Engineering Geology Bulletin, 52, 75–78, 1995.